Utilization of Mel-Spectrogram for Identification and Analysis of Animal Communication Patterns Acously

Abstract

This study aims to analyze the vocalization characteristics of eight animal species through Mel-spectrogram representation to identify dominant frequency patterns, duration, and sound intensity. Data were obtained from natural recordings and then converted into Mel-spectrograms, which were then analyzed to obtain information on frequency distribution, vocalization duration, and sound intensity levels. The results showed that most species showed dominant frequencies in the range of 2,000–9,000 Hz, with birds and chicks having the highest range (4,000–9,000 Hz). Vocalization patterns tended to be repetitive with varying durations (1–27 seconds) and high intensities (0–20 dB) at the dominant frequencies. Intra-species variations were found in birds and cats, indicating the influence of age or social factors. These findings demonstrate the potential of Mel-spectrograms as a tool for analyzing animal vocalizations for conservation purposes, behavioral monitoring, and the development of artificial intelligence-based speech recognition systems.